Substantial improvements in large-scale redocking and screening using the novel HYDE scoring function.

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Scoring functions are crucial for predicting protein-ligand binding affinity.
• Existing functions often rely on affinity data, limiting their general applicability.
• There is a need for broadly applicable scoring functions that provide interpretable results.

Purpose of the Study:

• To introduce and validate the HYDE scoring function for protein-ligand binding prediction.
• To demonstrate HYDE's applicability across diverse protein targets without using affinity data for calibration.
• To highlight the visualization capabilities of HYDE for analyzing binding interactions.

Main Methods:

• Developed the HYDE scoring function based on HYdration and DEsolvation terms, calibrated using octanol/water partition coefficients.
• Applied HYDE in large-scale redocking experiments on the Astex diverse set.
• Evaluated HYDE's performance in virtual screening using the DUD dataset.

Main Results:

• Achieved 93% accuracy in predicting correct binding modes during redocking calculations.
• Demonstrated significant enrichment in virtual screening with a mean AUC of 0.77.
• The atom-based score provides intuitive visualization of protein-ligand interactions.

Conclusions:

• HYDE is a generally applicable and accurate scoring function for protein-ligand binding.
• Its intuitive scoring facilitates lead optimization by visualizing binding contributions.
• Identified potential pitfalls in current benchmark datasets for future improvements.